NARROW FACETS OF HONESTY-HUMILITY PREDICT COLLEGIATE CHEATING
1As collegiate cheating is a growing concern across the world, researchers have started to explore personality traits for a better understanding of why students engage in such behavior.
While prior studies investigated the relationship between personality and general academic dishonesty criteria (viz., counter-academic behavior), this study aimed to provide a deeper understanding of how narrow facets of honesty-humility (i.e., HEXACO Personality Inventory) predict both general and specific academic dishonest behavior. Results from our study of 308 university students replicated prior findings in that the narrow facet, fairness, was the best predictor of broad counter-academic behavior. However, our data revealed that greed avoidance predicted a more specific form of academic dishonest behavior, namely collegiate cheating. Furthermore, narrow facets of honesty-humility show incremental predictive validity above and beyond the global trait when explaining variance in both general and specific self- reported measures.
1 A version of this work is published in: Janse van Rensburg, Y-E., De Kock, F.S., & Derous, E. (2018). "Narrow facets of honesty-humility predict collegiate cheating." Personality and Individual Differences 123, 199-204.
https://doi.org/10.1016/j.paid.2017.11.006
A previous version of the study was presented at The European Association for Work and Organizational Psychology, Dublin, Ireland, May 2017.
Introduction
Internationally, collegiate cheating has become a central concern for educational stakeholders (Bretag, 2016; Gallant, Binkin, & Donohue, 2015). A review of 64 studies of general cheating prevalence (amongst United States and Canadian college students) revealed an average of 43.1% students confessing to having cheated on examinations in some way (Whitley, 1998). More recent research – a longitudinal study including data from 134 709 students over the period 2002 to 2013 – reports an average of 37.3% students cheated by receiving unauthorized help whilst conducting written tasks (McCabe, 2016). As collegiate cheating appears to be a prevailing problem, a more comprehensive understanding of its psychological underpinnings is needed.
Prior studies (Marcus et al., 2007; McAbee, Oswald, & Connelly, 2014) show that students’ personality traits might play an important role in academic dishonest behavior. For example, honesty-humility (one of the six broad traits derived from HEXACO-Personality Inventory; Lee & Ashton, 2004) has consistently shown negative relationships with counter- academic behavior (De Vries et al., 2011). Counter-academic behavior is defined as a broad criterion, encapsulating multiple ethical transgressions within an academic context (i.e., cheating, plagiarism, substance abuse, misrepresentation, low personal standards, petty personal gain; Hakstian, Farrell, & Tweed, 2002). Remarkably, we do not yet understand how honesty-humility might relate to more context-specific measures of counter-academic behavior, such as cheating on a test. Counter academic behavior encompasses various types of wrongdoing, whereas the latter accounts for a specific type of dishonest behavior. Given that collegiate cheating represents a serious violation in the academic context (Bretag, 2016; Gallant et al., 2015), it is important to determine whether honesty-humility predicts this specific form of counter-academic behavior.
The present study investigated how narrow facets of honesty-humility relate to different types of academic dishonesty criteria (i.e., differentiated by generality insofar that counter- academic behavior includes several, varied outcomes whereas collegiate cheating is more specific). Although it has been investigated how personality factors relate to counter-academic behavior as general criterion, how personality predicts cheating as a more specific form of dishonest academic behavior has not yet been explored. Therefore, we sought to replicate the findings of earlier studies (De Vries et al., 2011) and establish how narrow facets of honesty- humility relate to collegiate cheating.
Another issue to consider was how well collegiate cheating could be predicted from narrow facets (versus the global trait) of honesty-humility, in the light of predictor specificity.
In the personality literature (Soto & John, 2017), it is suggested that narrow facets may predict context-specific behavior (e.g., collegiate cheating) with higher accuracy compared to global traits. Therefore, a secondary aim was to establish whether narrow facets of honesty-humility are better predictors of collegiate cheating compared to the global trait measure.
Honesty-Humility and Counter-Academic Behavior
Honesty-humility is the “quintessential basic trait to account for individual differences in ethical behavior” (Hilbig & Zettler, 2015, p. 85). This personality trait is defined as “the tendency to be fair and genuine in dealing with others” (Ashton & Lee, 2007, p. 156). Honesty- humility as a global trait comprises two theoretical aspects: honesty and humility (Leone et al., 2012). Honesty is defined as “the refusal to pretend that facts of reality are other than what they are” (Becker, 1998, p. 158). Put differently, honesty refers to being fair and trustworthy, unwilling to cheat, exploit, steal or lie to others for personal gain (Ashton et al., 2000).
Humility, on the other hand, is about being modest and avoiding greed. Individuals high on humility are uninterested in acquiring luxury goods or having high social status (Ashton & Lee, 2001, 2002).
Further, the global trait of honesty-humility is operationalized as comprising four narrow facets (Lee & Ashton, 2004). Together, the two narrow facets of sincerity (being genuine in interpersonal relations) and fairness (avoiding fraud and corruption) theoretically represent the element of honesty. It is specifically suggested that people high on honesty have the internal control to avoid fraud, stealing or cheating, irrespective of their motivation in wanting to do so (Zettler & Hilbig, 2010). As individuals high on honesty are associated with fostering relationships based on trust and reciprocity, one would expect less counter-productive behavior from such individuals. In contrast, the two narrow facets of greed avoidance (being uninterested in possessing lavish wealth, acquiring luxury goods and having high social status) and modesty (viewing oneself as an ordinary person, unentitled to special treatment), represent the aspect of humility. A person high on humility would have less motivation to compete for acquiring social status or material advantages (Leone et al., 2012) as humility represents an accurate view of one’s actual traits, abilities or resources. To summarize, greed avoidance and modesty could be the underlying motive for engaging in unethical behavior; and fairness and sincerity may – despite the motive – act as a control element, buffering whether one would actually engage in counter-productive behavior.
Counter-productive behavior, specifically conducted within an academic or collegiate setting, is referred to as counter-academic behavior (Hakstian et al., 2002). Counter-academic
behavior has been operationalized as self-proclaimed ethical violations and is computed as the mean score of various counter-productive behaviors added together (e.g., misrepresentation, cheating during examinations, plagiarizing, etc.). Counter-academic behavior has been found to relate negatively to the global trait of honesty-humility (r = -.40, p < .01; showing medium to large effect) (De Vries et al., 2011). However, the narrow facets underlying honesty-humility may predict academic dishonesty differently; the reason being ascribed to the conceptual difference of facets and the way counter-academic behavior is conceptualized.
The narrow facets of honesty-humility may relate to counter-academic behavior in different ways. For instance, De Vries et al. (2011) found that the four narrow facets of honesty- humility correlated negatively with broad counter-academic behavior. In their research, they explored how all six global traits of the HEXACO-PI relate to counter-academic behavior.
More relevant to our research, De Vries et al. (2011) found that, although all narrow facets of honesty-humility significantly correlated with counter-academic behavior, only one narrow facet, namely fairness, predicted counter-academic behavior (β = -.47 p < .01). Further, the findings of De Vries et al. (2011) revealed that fairness (which is expressed as relative weight in percentages) explained 67.7% of the total variance in predicting counter-academic behavior.
Following De Vries et al. (2011), we expected to replicate the findings that:
Hypothesis 1. The narrow facets of fairness (H1a), sincerity (H1b), greed avoidance (H1c) and modesty (H1d) will relate negatively to counter-academic behavior, with fairness relating more negatively to counter-academic behavior than sincerity, greed avoidance and modesty (H1e).
Honesty-Humility and Collegiate Cheating
Many studies have so far failed to consider how the narrow facets of honesty-humility would relate to a context-specific measure of counter-academic behavior, like cheating. As a consequence, we do not yet understand how honesty-humility, at facet-level, might relate to collegiate cheating (Hilbig & Hessler, 2013).
In prior studies (De Vries et al., 2011; Marcus et al., 2007; McAbee et al., 2014) counter-academic behavior was defined as a broad form of academic dishonesty, encapsulating multiple ethical transgressions like plagiarizing, or cheating. In contrast, it is also possible to conceptualize academic dishonesty in a narrower way, fitting a particular situational context.
For example, cheating during a test could be considered a context-specific criterion measure of academic dishonest behavior.
Recently, researchers have found a significant relationship between the global trait, honesty-humility, and the probability of actual cheating under monetary incentivized conditions (Hilbig & Zettler, 2015). For example, one study (Zettler et al., 2015) reports that low scores on honesty-humility significantly relate to over-reporting on alleged wins during a monetary incentivized coin-toss task. Closer to the notion of using narrow facets as predictors of criteria, Hilbig, Glöckner, and Zettler (2014) predicted that honesty-humility would explain unique variance in predicting prosocial behavior (being the opposite of self-interest behavior such as cheating) due to the inclusion of the fairness and greed avoidance aspect. Although both these narrow facets were significantly associated with prosocial behavior, only fairness explained unique variance in this outcome. However, how honesty-humility, at facet level, might relate to collegiate cheating has not been tested yet.
As far as current literature reports, a person high on fairness would want to avoid fraud, corruption, stealing and cheating (Lee & Ashton, 2004; Zettler & Hilbig, 2010). Further, fairness has been found to be the best predictor of counter-academic behavior when compared to the other three narrow facets of honesty-humility (De Vries et al., 2011). Furthermore, counter-academic behavior was found to include aspects of self-reported cheating during examination (captured by items in the counter-academic behavior measure; Hakstian et al., 2002). Therefore, we expected that students that are low on fairness would also be more likely to engage in dishonest behavior like cheating. Hence our premise:
Hypothesis 2. The narrow facets of fairness (H2a), sincerity (H2b), greed avoidance (H2c) and modesty (H2d) will relate negatively to collegiate cheating, with fairness relating more negatively to collegiate cheating than sincerity, greed avoidance and modesty (H2e).
Researchers have demonstrated that strong relations and increased validity may result when narrow-level facets of personality constructs are related to specific criteria, as opposed to using the global trait alone (Barrick & Mount, 2003). To increase predictive precision, a facet-level research approach could be followed when studies pose context‐ specific research questions and wish to predict specific outcomes (Judge & Kammeyer‐ Mueller, 2012; Soto &
John, 2017). This notion suggests that the narrow facets of honesty-humility could show incremental predictive validity over and above that of the global trait, honesty-humility, when
predicting specific dishonest behavior (viz., collegiate cheating). Such facet-level insights are important, because they may provide a better theoretical understanding of how personality and criteria are linked (Hastings & O’Neill, 2009).
In fact, prior studies using a broad academic counter-behavior criterion (De Vries et al., 2011) showed that a model containing the two global traits, honesty-humility and conscientiousness (R2 =.29, p < .01) versus a model containing their narrow facets (R2 = .41, p < .01) respectively, incrementally predicted counter-academic behavior (ΔR2 = .12, p < .01).
Thus, by using narrow facets, rather than the global trait predictors, an additional 12% of variance in counter-academic behavior was explained. Therefore, to establish whether narrow facets of honesty-humility are better predictors of collegiate cheating compared to the global trait measure, we proposed:
Hypothesis 3. The set of honesty-humility narrow facets (fairness, sincerity, greed avoidance, modesty) will explain more variance in collegiate cheating than a model containing only the global trait (honesty-humility).
Method Participants and Procedure
An email was sent to 700 students studying at a large South African university, requesting them to participate in our study. Of these, 392 students started the survey, but either did not complete the questionnaire or did not give consent after the purpose of the research had been made known. Therefore, the final sample comprised 308 participants who completed the online questionnaire. The mean age of the sample was 22.9 years (SD = 5.23; ranging from 18 to 47 years), comprising diverse ethnic backgrounds (Black = 37.3%, White = 37.7%, Indian
= 12%; mixed Black/White, = 6.5%, Asian = 2.3%, other = 4.2%). The sample represented 64.7% bachelor students (equally spread across the three years) and 35.3% postgraduate students (honors, master’s, PhD). Men and women were almost equally represented (157 women; 151 men).
After obtaining their informed consent, participants commenced with the questionnaire measuring variables in the following sequence: demographics, collegiate cheating (disguised as a cognitive task), honesty-humility, and counter-academic behavior. Next, we debriefed participants by disclosing that the “cognitive task” elicited cheating behavior. Finally, participants were given the opportunity to self-report whether they had cheated or not (see
4.2.3). Participants then had the option to voluntarily withdraw their consent. Before conducting this study, approval was received from ethics board of the relevant university.
Measures
HEXACO-PI. Honesty-humility was measured with sixteen items of the honesty- humility scale of the HEXACO-PI, which is a widely used and valid personality questionnaire (Ashton, Lee, & De Vries, 2014; Lee & Ashton, 2004). According to the literature, honesty- humility (global trait) consists of four narrow facets, captured by four items respectively. All items are measured on a 5-point Likert-type scale (1 = strongly disagree to 5 = strongly agree).
Each of the narrow facets, namely fairness (α = .71), greed avoidance (α = .76), modesty (α = .60) and sincerity (α = .60) showed acceptable internal consistency.
Counter-academic behavior. Self-reported counter-academic behavior was captured by using the Inventory of Counter-productive Behavior (ICB; Hakstian et al., 2002). Like others (De Vries et al., 2011; Marcus et al., 2007), 25 items measuring academic counter- productive behaviors were measured on a 6-point Likert-type scale (1 = never even considered it, to 6 = did it three or more times). Example items include submitted a class paper or project that was not your own work (misrepresentation); and turned in work that was of poor quality and lower than your true potential or ability (low personal standards). The measure showed good (α = .86) internal consistency.
Collegiate cheating measure. Building on previous studies (Halevy, Shalvi, &
Verschuere, 2014; Peer, Acquisti, & Shalvi, 2014), we measured students’ self-reported cheating on a task which was presented as a cognitive ability test.2 In a standardized situation, respondents had the opportunity to win money if they performed well, thus making cheating an attractive and non-risky option.3 During the task, explicit instructions were given not to use any unauthorized help (i.e., using the internet, a calculator or asking help from a friend) or to change answers whilst looking at the answer sheet, as this would constitute cheating.
Participants scored their own answers against an electronic answer sheet. Following Halevy et al. (2014), self-reported cheating was captured as a binary outcome (yes I cheated = 1, or no, I did not cheat = 0), irrespective of the method or frequency of cheating (Peer et al., 2014).
2 See Appendix B: Information about the cheating task. To view items used in the cheating task, see Chapter 5, Appendix E: Cheating Measure Instructions to Respondents.
3 We acknowledge an anonymous reviewer for wording of the text.
Collegiate cheating was therefore contextualized to capture self-reported cheating about a specific situation that had been experienced immediately beforehand, making this criterion more precise as compared to general counter-academic behavior.
Results
Before the main analyses were conducted, we checked missing values (with no missing data), various assumptions, descriptive statistics and correlations (Table 3.1). Next, in following best-practice recommendations on whether to include/exclude control variables, we applied the decision-making tree by Bernerth and Aguinis (2016), answering a series of questions that summarize sequential steps in the process of selecting control variables.
According Bernerth and Aguinis (2016) researchers too often include unwarranted control variables. We first considered empirical findings about the relation between collegiate cheating, gender and age. To date there are non-consistent results about age and gender differences with regards to cheating in academic settings (Crown & Spiller, 1998; Gallant et al., 2015). Since there is no strong theoretical rationale about the relationship between the focal variable (cheating) and control variables (age and gender), we elected to not include any control variables.
Table 3.1
Descriptive Statistics and Intercorrelations of Study Variables
Variable M SD 1 2 3 4 5 6 7 8
1. aGender – –
2. Age 22.94 5.23 .03 –
3. Honesty-humility 3.50 0.60 -.21** .18** α = .83
4. Fairness 3.84 0.89 -.25** .16** .78** α = .71
5. Sincerity 3.44 0.75 -.05 .08 .67** .39** α = .60
6. Greed avoidance 2.99 0.87 -.10 .16** .73** .39** .28** α = .75
7. Modesty 3.71 0.75 -.20 .12* .75** .47** .35** .43** α = .60 8. bCAB 2.43 0.71 .18** -.25** -.42** -.48** -.27** -.26** -.19** α = .86 9. cCollegiate cheating 0.32 0.47 -.04 -.20** -.11* -.04 -.09 -.15** -.06 .12* Notes. N = 308 of which n = 99 cheated on the task. Internal consistencies (α) are available on the diagonal.
aGender was coded 0 = female and 1 = male. bCAB = counter-academic behavior. cCollegiate cheating was coded as 1 = did cheat and 0 = did not cheat; cheating (as a binary variable) and other continuous variables were correlated using Point Biserial correlation.
*p < .05. ** p <.01 (two-tailed).
Our results were in support of Hypothesis 1, as fairness did show the strongest negative relationship with counter-academic behavior (H1a; r = -.48, p < .01; medium to strong effect size), compared to the relationships of the other narrow facets: sincerity (H1b; r = -.27, p <
.01); greed avoidance (H1c; r = -.26, p < .01) and modesty (H1d; r = -.19, p < .01). The Steiger Z-test (Steiger, 1980) revealed that the relationship between fairness and counter-academic behavior was indeed significantly stronger (H1e) than the relationship between sincerity and counter-academic behavior (Z = -3.72, p < .001), greed avoidance and counter-academic behavior (Z = -3.89, p < .001), and modesty and counter-academic behavior (Z = -5.42, p <
.001).
Hypothesis 2 was not supported by our data, because the relationship between collegiate cheating and fairness (H2a; r = -.04, p > .05), sincerity (H2b; r = -.09, p >.05) and modesty (H2d; r = -.06, p >.05) were all non-significant. Surprisingly, only greed avoidance significantly related to collegiate cheating (H2c; r = -.15, p < .01; small effect size). For this reason, we only tested whether greed avoidance related more negatively to cheating than fairness (H2e). No significant difference was found in comparing the relationship between greed avoidance and cheating versus fairness and cheating (Z = -1.75, p >.05).
Finally, Hypothesis 3 explored whether the set of narrow facets of honesty-humility would explain more variance in collegiate cheating than a model containing only the global trait. To test the predictive power of the narrow facts, binary logistic regression and relative weight analyses were conducted (Tonidandel & LeBreton, 2015). First, cheating was regressed onto the global trait honesty-humility (see Table 3.2, Model 1; R2 = .018, p < .05), and then onto the narrow facets: fairness, sincerity, modesty and greed avoidance (Model 2; R2 = .035, p < .05). Only greed avoidance was significant in predicting the odds of cheating, namely Exp(β) = .69, χ²(8) = 6, p < .05 (Model 2), implying that as greed avoidance increases, the odds of cheating decreases. The odds ratio corresponds with a small to medium-sized effect (Rosenthal, 1996). Hypothesis 3 was confirmed in that the set of narrow facets for honesty- humility incremented the prediction over a model containing the global trait (ΔR2 = .017, p <
.05). The result had a small effect, f2 = .02 (Cohen, 1988).
Table 3.2
Logistic Regression and Relative Weights Analysis of Collegiate Cheating on Honesty- Humility as a Global Trait and the Narrow Facets
We further explored whether the set of narrow facets of honesty-humility would also – as in the case of collegiate cheating – explain more variance in counter-academic behavior than the global trait. By regressing counter-academic behavior on honesty-humility as a global trait (Step 1) and narrow facets (Step 2), the same pattern emerged as narrow facets showed incremental predictive validity in predicting counter-academic behavior. Additionally, we also obtained the relative weights (RW) for hierarchical regression.
Predictor Collegiate cheating
β Exp(β) [95% CIa] p RWraw [95%CI] RW%b
Model 1 (global trait):
Honesty-humility -.40 .67* [-.850; -.023] .05
-2 log likelihood 382.95
R² (N) .02*
R²(CS) .01*
Model 2 (narrow facet):
Fairness .10 1.11 [-.223;.461] .53 .000 [2.518; 0.001] 3.14%
Sincerity -.19 .83 [-.559; .180] .31 .005 [2.799; 0.027] 19.01%
Modesty .02 1.02 [-.429; 476] .93 .001 [1.663; 0.003] 5.53%
Greed avoidance -.37 .69* [-.715; -.078] .02 .018 [1.020; 0.053] 72.31%
-2 log likelihood 378.95
R² (N) .04*
R²(CS) .03*
Notes. N = 308. R² (N) = Nagelkerke; R² (CS) = Cox & Snell; cheating was coded as 0 = did not cheat, and 1 = did cheat, so positive β values indicate positive associations with cheating.
Hosmer and Lemeshow Test for Model 1 (Honesty-humility global trait): χ²(8) = 10.2, p > .05. Model 2 (narrow facets): χ²(8) = 6, p > .05.
a95% confidence interval (95%). Bootstrapped results are based on 1 000 bootstrap samples.
bRelative weights are not raw weights but rescaled to express the % contribution of each predictor to overall R2. Confidence interval around the raw weights was calculated using the bias corrected accelerated method for generating the bootstrapped confidence intervals.
*p < .05. ** p < .01.